Equity Wall Strategy Research

Gate Metrics

Net Sharpe vs Turnover

Research Report

# Equity Wall Strategy Research Report

## Executive conclusion

Do not promote to paper trading yet. The observable equity-wall idea is plausible as a diversifier, but this first implementation is not strong enough for PAW paper deployment.

- Baseline SPY gate: net Sharpe 0.875, net CAGR +14.15%, max DD -33.72%, but P0 bias gate failed and factor R2 is 0.960; treat it as mostly SPY beta.
- Robust rotation: bias gate passed, net Sharpe 0.542, net CAGR +3.21%, max DD -12.58%, OOS Sharpe 1.059, max corr to supplied PAW proxies 0.270.
- Robust blocker: turnover 7.206x/year, TC drag 0.728% NAV/year, 252d rolling Sharpe min -2.276, and negative rolling windows +29.84%.

## Observable equity-wall definition

An equity wall is defined as broad, synchronized buying pressure in liquid equity ETFs. The proxy uses only public adjusted close and volume:

- `signed_flow = daily_return * log1p(dollar_volume / ADV63)`.
- `flow_z = trailing_126d_zscore(signed_flow)`, clipped to [-5, 5].
- Aggregate wall score is the mean core broad-market ETF `flow_z` across SPY, QQQ, IWM, DIA, VTI, and RSP.
- Positions on date `t` use signals through previous trading day `t-1`.

## Candidate universe and data sources

- Traded universe: SPY, QQQ, IWM, DIA, VTI, RSP, MTUM, QUAL, VLUE, USMV, XLK, XLY, XLF, XLI, XLE, XLV, XLP, XLU, XLB, XLRE, EFA, EEM.
- Factor proxies: SPY, QQQ, IWM, MTUM, QUAL, VLUE, USMV, EFA, EEM, TLT, GLD, UUP.
- Data source: Yahoo Finance via `yfinance`, adjusted daily close and volume, 2010-01-04 through 2026-06-08.
- Existing-return comparison: local PAW research/live-eval gross return streams where available plus SPY/QQQ/IWM benchmarks; this is not actual live P&L.

## Candidate definitions

- Simple baseline: `equity_wall_spy_gate_v1`; hold 100% SPY when aggregate wall > 0.25 and SPY 63d momentum > 0, otherwise cash.
- Robust variant: `equity_wall_rotation_robust_v1`; when aggregate wall > 0.10 and SPY 63d momentum > 0, select up to five ETFs by 60% flow z-score plus 40% 63d volatility-adjusted momentum, inverse-vol weight capped at 35%, and move half-way to target weekly.

## Eval table

| Candidate | Bias | Net Sharpe | Gross Sharpe | Net CAGR | Net Total | Max DD | OOS Sharpe | Turnover | TC drag | Factor R2 | Max corr |
|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
| equity_wall_spy_gate_v1 | False | 0.875 | 0.875 | +14.15% | +747.89% | -33.72% | 1.433 | 0.031 | 0.003% | 0.960 | 0.565 |
| equity_wall_rotation_robust_v1 | True | 0.542 | 0.660 | +3.21% | +66.56% | -12.58% | 1.059 | 7.206 | 0.728% | 0.292 | 0.270 |

## Gate decision

- P0: baseline fails due to eval-kit shift-lookahead test; robust passes causality but has only moderate net performance.
- P1: robust has diversification merit with max abs correlation 0.270 to supplied proxies, but rolling stability and transaction-cost sensitivity are not good enough.
- P2: not live-ready; paper/live return stream, fill assumptions, queue/latency handling, and true existing-strategy P&L are missing.

## Paper-trading recommendation

No paper trade now. Iterate only if a lower-turnover robust variant can reach net Sharpe above 0.8, preserve max drawdown below 15%, keep max correlation below 0.8 to actual PAW strategy P&L, and pass DSR/PBO on a candidate trial panel.